By Steven Martin
Chief Commercial Officer, GE Digital & Chief Digital Officer, GE Power at GE
There was a lot of dialogue surrounding this article. Not surprising, given the increasingly large pool of data on which to run machine learning algorithms and growing interest in Industrial IOT.
To continue the conversation, I’ll offer four predictions on how the Industrial IoT market will play out over 2019…
#1 Aggregated data will beat algorithms (even really good ones) every time
One thing we have learned at GE from decades of creating industrial technology is that the ability to securely aggregate data across an entire fleet offers the best insights – better than any device-specific algorithm could on its best day.
As Peter Norvig, Director of Research at Google, is often quoted as saying, “We don’t have better algorithms. We just have more data.” More is just better, as long as you know what to do with it.
#2 Domain knowledge is irreplaceable
Did you know that on average, less than 5% of all industrial data is reviewed? It’s not that it’s unusable. It requires very specific expertise to consume it and gain insights. Those of us closest to the hardware are best positioned to optimize it. Why? Because we have knowledge of the asset and ability to identify which data sets are most integral for optimization.
The hype around independence from system manufacturers is just that. Hype. Even the brightest Data Scientists cannot make up for a lack of hardware insight and access to aggregated data.
Know the asset if you want to push the optimization boundaries.
#3 Cloud providers will move higher in the stack
New? No. Revolutionary? Not that either…
It pains me that those of us in industrial domains have to remind ourselves of this reality. Cloud vendors will offer a lot of what is needed for Industrial IoT platforms. They will get better and better at it.
A truth since the dawn of the software industry: Pure-play vendors get consumed by larger infrastructure providers reducing risk and complexity. They will do this across data ingestion, data management, analytics and a ton of other places.
Scary? Maybe, but no more so than driving to work because of its unpredictability. It’s going to happen. If you need it, here’s a mantra that might bring you inner peace:
This is something we want.
This is something we expect.
This is something we (enthusiastically) embrace.
If this is not something I want, expect or embrace I am on the wrong side of history.
Any company attempting to sit between system experts and the cloud providers driving the infrastructure will feel the pressure of this shift. In short, “pure-play” providers will find themselves in a suicide squeeze between Carlton Fisk and George Brett while pulling a training sled. Good luck. It won’t be pretty or close.
#4 The days of recommendations are limited, very limited
At GE Power we increasingly find ourselves aligning APM and OPM functionality much more closely to our operational systems so that action is immediately taken. Running a report on things we can do to tune a system is helpful, but not nearly as helpful as automating that system to implement the change (within safety boundaries of course).
We know how because of the vast expertise accumulated by the GE employees who have spent their entire careers building and optimizing hardware alongside our valued customers.
Revolutionary hardware deserves revolutionary software. If you only have one or the other, you actually have neither.